Original Guideline: December 2002
Trials of preoperative radiotherapy (RT) versus surgery alone were pooled using the software package Metaanalyst0.998 (J. Lau, Boston, MA, USA). Overall mortality, local failure, tumour resectability, tumour downstaging, and adverse effects were pooled in separate analyses for all studies, where data was available. Reported figures or estimates obtained from tables or graphs were used. For calculation of survival and local failure, all eligible patients were considered in the denominator, based on intention to treat. All deaths at the time of reporting, regardless of cause, were included in survival calculations. Patients with local failure included those with non-resected as well as those with recurrent disease. Only resected cases were considered in the calculation of downstaging.
Data were pooled using the random effects model as the more conservative estimate of effect. Results were expressed as risk ratios (RR) with 95% confidence intervals (CI), where a RR less than 1.0 favours preoperative RT and a RR greater than 1.0 favours surgery alone. Odds ratios (OR) and absolute risk differences (RD) were also calculated.
Heterogeneity of results among trials was expected in view of the different treatments used and populations tested, as well as the wide time interval and geography across which these trials were conducted. For example, the RT prescription may affect the results. RT doses greater than 30 Gy10 are considered necessary and pelvic fields are as effective as extended fields. Moreover, the use of three or more RT beams will lessen toxicity, and short delays of surgery after RT will not demonstrate downstaging. Thus, these factors were investigated with sensitivity analyses to see whether there was an impact on results. Outcomes of predetermined groups of patients were examined initially by the graphic method described by L'Abbe et al. and RR calculated. For sensitivity analyses the following factors were examined:
Treatment effects:
- Biologically effective dose (BED) of RT (less than 30 Gy10 versus equal to or greater than 30 Gy10). BED was calculated using the formula BED=nd (1+d/alpha/beta), where n=number of fractions, d=dose per fraction, alpha/beta=10 for tumour effect and acute reactions and alpha/beta=3 for late reactions, with no time correction (not needed for late reactions) because parameters were not available and usual ranges are quite wide;
- RT fraction size (standard fractions up to 2.5 Gy/day versus high fractions of 5 Gy/day);
- Contemporary radiotherapy prescription, defined as studies employing multiple-field technique and target volume confined to the pelvis (i.e., excluding studies employing parallel pair arrangements or including para-aortics); and
- Delay of surgery after completion of RT (less than seven days versus eight or more days).
Population effects:
- Studies including a range of rectal cancer cases versus those including only advanced disease.
Sensitivity analyses were also performed for all five of the meta-analyses (overall survival, local failure, tumour resectability, downstaging, and adverse effects) considering only trials with high design quality. The quality of the 14 eligible randomized trials of preoperative RT versus surgery alone in operable rectal cancer was scored independently. Five assessors assessed each trial using the Detsky instrument. This questionnaire addresses five domains of study quality: randomization process, outcomes measure, patient eligibility, treatment description, and statistical procedures. The 14 questions on the Detsky instrument can be answered "adequate," "inadequate," or "partial" and scored 1, 0, or 0.5, respectively. The final score of each trial is a ratio of the observed points divided by the total number of questions answered. The results from the five assessors were averaged for a final score. Trials with Detsky instrument scores greater than 0.5 were considered to be of high quality.
January 2004 Update
Where the Metaanalyst0.998 (Dr. Joseph Lau, Boston, MA, USA) software program was used to perform all meta-analyses and produce all figures in the original guideline, in the manuscript and this update version, Review Manager 4.2.1 (© Update Software) (which is freely available through the Cochrane Collaboration) was used. All figures and tables in this practice guideline have been updated to reflect the latest information as presented in the manuscript.
At the suggestion of one of the peer reviewers, the first bullet under treatment effects was changed to include a correction for time and now reads as follows in both the manuscript and this update:
Biologically effective dose (BED) of RT (less than 30 Gy10 versus equal to or greater than 30 Gy10). BED was calculated using the Linear Quadratic formula and the parameters suggested for time correction (1u):
BED time = nd (1+d/alpha/beta) - gamma/alpha (T – Tk) where n=number of fractions, d=dose per fraction, alpha/beta=10 for tumour effect and acute reactions and alpha/beta=3 for late reactions, gamma/alpha = repair rate set at 0.6 Gy/day and T = total treatment time and Tk = initial delay time set at 7 days.